Clim. Past, 3, 109–118, 2007 www.clim-past.net/3/109/2007/ Climate © Author(s) 2007. This work is licensed of the Past under a Creative Commons License. Predicting Pleistocene climate from vegetation in North America C. Loehle National Council for Air and Stream Improvement, Inc., 552 S Washington St., #224, Naperville, IL 60540, USA Received: 22 September 2006 – Published in Clim. Past Discuss.: 23 October 2006 Revised: 8 January 2007 – Accepted: 8 February 2007 – Published: 12 February 2007 Abstract. Climates at the Last Glacial Maximum have been 1 Introduction inferred from fossil pollen assemblages, but these inferred climates are colder for eastern North America than those pro- The Pleistocene Last Glacial Maximum (LGM) period of duced by climate simulations. It has been suggested that low 18 000 years ago has been widely interpreted as a time of CO2 levels could account for this discrepancy. In this study bitter cold in eastern North America when tundra and bo- biogeographic evidence is used to test the CO2 effect model. real forest extended hundreds of miles south of the ice sheets The recolonization of glaciated zones in eastern North Amer- and the temperate forest of the East retreated to the south- ica following the last ice age produced distinct biogeographic ern coastal plain, to Florida, and westward into Texas and patterns. It has been assumed that a wide zone south of the Mexico (Davis, 1983, 1984; Davis and Shaw, 2001; Deevey, ice was tundra or boreal parkland (Boreal-Parkland Zone or 1949; Delcourt and Delcourt, 1984, 1993; Jacobson et al., BPZ), which would have been recolonized from southern 1987; Maher et al., 1998; Maxwell and Davis, 1972; Over- refugia as the ice melted, but the patterns in this zone dif- peck et al., 1992; Prentice et al., 1991; Ritchie, 1987; Royall fer from those in the glaciated zone, which creates a major et al., 1991; Schoonmaker and Foster, 1991; Tallis, 1991; biogeographic anomaly. In the glacial zone, there are few en- Watts, 1970, 1971, 1973, 1979, 1980a, b; Watts and Stuvier, demics but in the BPZ there are many across multiple taxa. 1980; Webb et al., 1988, 1993; Whitehead, 1973; Wilkins In the glacial zone, there are the expected gradients of ge- et al., 1991). This reconstruction, which may be called the netic diversity with distance from the ice-free zone, but no standard model, is commonly presented in textbooks (e.g., evidence of this is found in the BPZ. Many races and re- Bradley, 1999; Delcourt and Delcourt, 1993; Pielou, 1991; lated species exist in the BPZ which would have merged or Ritchie, 1987; Tallis, 1991). The standard model is based hybridized if confined to the same refugia. Evidence for dis- largely on pollen and macrofossil records. These pollen tinct southern refugia for most temperate species is lacking. data have been interpreted qualitatively in some cases, and Extinctions of temperate flora were rare. The interpretation in other cases transfer functions or response surface models of spruce as a boreal climate indicator may be mistaken over (e.g., Farrera et al., 1999; Nakagawa et al., 2002; Peyron et much of the region if the spruce was actually an extinct tem- al., 1998; Tarasov et al., 1999; Webb et al., 1993, 1997) have perate species. All of these anomalies call into question the been used to infer climate from pollen composition, with concept that climates in the zone south of the ice were ex- similar results. By any of these three methods, the inferred tremely cold or that temperate species had to migrate far to LGM climate is much colder than that simulated (Ganopol- the south. An alternate hypothesis is that low CO2 levels gave ski et al., 1998; Huntley et al., 2003; Kageyama et al., 2001; an advantage to pine and spruce, which are the dominant Pinot et al., 1999; Webb et al., 1993, 1997). Determining the trees in the BPZ, and to herbaceous species over trees, which correct paleotemperature is important for calibrating general also fits the observed pattern. Thus climate reconstruction circulation models (e.g., Crowley, 2000; Farrera et al., 1999; from pollen data is probably biased and needs to incorporate Ganopolski et al., 1998; Pinot et al., 1999). In addition, the CO2 effects. Most temperate species could have survived vegetation composition at this time is consistently described across their current ranges at lower abundance by retreating as having no analogs in modern periods. In this paper it is to moist microsites. These would be microrefugia not easily argued that both of these anomalies arise from a combination detected by pollen records, especially if most species became of ambiguous pollen interpretation and the effects of low am- rare. These results mean that climate reconstructions based bient CO2 at the LGM which would have altered the relative on terrestrial plant indicators will not be valid for periods dominance of different taxa in a manner that mimics colder with markedly different CO2 levels. and drier climates. Biogeographic and phylogenetic data are cited as independent tests of the CO2 effect model. Correspondence to: C. Loehle The idea that glacial levels of CO2 could affect vegeta- ([email protected]) tion structure and composition was proposed over a decade Published by Copernicus GmbH on behalf of the European Geosciences Union. 110 C. Loehle: Predicting Pleistocene climate ago (e.g., Beerling, 1996; Beerling and Woodward, 1993; found evidence for temperate forest vegetation right up to Bennet and Willis, 2000; Cole and Monger, 1994; Cowl- the ice margin throughout the period of the last glacial ad- ing, 1999; Cowling and Sykes, 1999; Farquhar, 1997; Harri- vance in Illinois, Indiana, and Ohio. Farther north and west, son and Prentice, 2003; Jolly and Haxeltine, 1997; Polley in Minnesota, Montana, and the Dakotas, the southern glacial et al., 1993, 1995; Street-Perrott et al., 1997). Neverthe- margins were frozen to their beds and signs of permafrost are less, LGM climate as estimated from pollen data generally present (Clayton et al., 2001; Mickelson et al., 1983). Pew´ e´ still does not factor in CO2 effects (e.g., Elenga et al., 2000; (1983) compiled abundant evidence for permafrost from the Nakagawa et al., 2002; Peyron et al., 1998; Tarasov et al., driftless area of Wisconsin and from across Pennsylvania, but 1999) and discussions of simulated LGM climate vs. vege- none from central Illinois east to the edge of Pennsylvania. tation (the anomaly problem) either do not mention the role Pew´ e´ (1983) shows some evidence for patterned ground on of CO2 (e.g., Ganopolski et al., 1998; Pinot et al., 1999) or the high peaks of the Appalachians, but only as far south as only touch on this effect (e.g., Crowley, 2000; Huntley et al., the southern Virginia-West Virginia border. Periglacial forms 2003). south of this either are undated or are not reliably identi- To date, the discussion has featured dueling models. Re- fied (Pew´ e,´ 1983). Denny (1951) documented frost action gression models based on pollen (which seem straightfor- in Pennsylvania that diminishes with distance from the ice ward) are pitted against simulations of plant communities margin. There is thus evidence for a band of permafrost and under low CO2. It is tempting to be wary of the simulations periglacial climate in Pennsylvania and Wisconsin but none because they cannot be directly verified. In what follows, in between. This reconstruction makes sense because the several types of independent evidence are presented that im- area of western Pennsylvania is higher in elevation than are ply that LGM climates inferred from pollen data are colder southern Ohio, Indiana, and Illinois. Thus, permafrost (and than the likely actual climates. These various types of ev- tundra) was not a universal feature of the ice front region in idence are shown to reconcile with effects of low ambient eastern North America as implied by most reconstructions. CO2, thereby supporting the models of CO2 effects. The re- Wider periglacial zones did of course occur in Europe. sults of the analysis also bear on questions of glacial refu- The lack of tundra puts limits on how cold this region gia. The focus of this study is largely eastern North America could have been. for the sake of concreteness, but no-analog vegetation found elsewhere at the LGM can be explained by similar mecha- nisms. 3 Taxonomic anomalies 2 Geologic anomalies A key species in the reconstruction of LGM climates in east- ern North America is spruce (Picea). The pollen of various By analogy with European mountains, it has been assumed species of spruce are very difficult to distinguish. Since all (e.g., Deevey, 1949; Delcourt and Delcourt, 1984) that the existing spruce species in eastern North America require a southern Appalachians, particularly the Great Smoky Moun- cold climate, they have generally been considered together tains, should have been covered by permanent ice caps and (see, e.g., Davis and Shaw, 2001). In this context, pollen should have generated glaciers. Mountain glaciers produce profiles dominated by spruce and herbaceous species have obvious signs like Ushaped valleys, striations, and moraines. been classified as boreal parkland. None of these signs has ever been found in the southern Ap- It is now apparent, however, that a major portion of the palachians south of Pennsylvania. More recent treatments southern range of spruce may have been occupied by a now- assume tundra but not glaciers on the peaks. extinct species (Jackson and Weng, 1999) which was com- Tundra soils produce distinctive signs, due to processes mon in the lower Mississippi valley and east at least to west- such as solifluction, cryoturbation, ice wedges, and stone ern Georgia. Given that fossil stumps of this species can be sorting.